DE102012215533A1 - Method for determining position of rail vehicle by satellite navigation system, involves providing receiver unit in retainer at track, and comparing recorded data in retainer with stored data of landmarks - Google Patents

Abstract

The method involves providing receiver units (105, 106) in a retainer at a track. Recorded data in the retainer is compared with stored data (104, 114) of landmarks. The landmarks are recognized to determine a position of a rail vehicle (109) correspondence to the retainer. One of the receiver units is arranged on a camera axis of the rail vehicle. Landmarks to-be-made are stored in the retainer during movement of the rail vehicle. The landmarks are building, stationary signal-technical device, signal, switch, tunnel entry, tunnel exit, building, railroad overpass, rock formation and bridge. Independent claims are also included for the following: (1) a satellite navigation system (2) a rail vehicle.

Description

The invention relates to a method for determining the position of a rail vehicle and to an associated training method. Furthermore, correspondingly designed rail vehicles and a system with at least one such rail vehicle are proposed.

A central challenge for command control is to determine the position of a rail vehicle in a route network with a high degree of safety or accuracy.

Wavelet transformation (WT) is a technique used to reduce image data. It is a decorrelation that removes redundant, irrelevant and unimportant image details from the image data. The transformation is a recursive process based on any number of iterations. The image content of the image to be transformed is analyzed with a high pass and a low pass. The two filter frequencies are chosen so that they halve the frequency range of the image (cf. http://www.itwissen.info/definition/lexikon/Wavelet-Transformation-WT-wavelet-transformation.html ).

"Prediction by Partial Matching" (PPM) is a family of adaptive statistical data compression algorithms built on context models and forecasting. PPM models use a set of symbols from the previous symbol stream to predict the next symbol of the stream. For example, predictions are limited to ratings of the symbols. The number of preceding symbols n determines the order of the PPM model, which is recorded as PPM (n). If no prediction can be made on the basis of all n context symbols, a prediction on the basis of n - 1 is attempted. This procedure is repeated until a hit is found or no symbols remain in the context. At this time, a prediction is set (cf. http://en.wikipedia.org/wiki/ Prediction_by_Partial_Matching ).

Current methods for determining the position of a rail vehicle in a route network are based on the reading of stationary brands (balises) and the balancing of the read balise with a route map. The reading in of a balise takes place with a high technical security. An uncertainty with which balises can be read incorrectly corresponds to the specified high technical safety requirements in railway traffic.

Furthermore, it is known to carry out the train position determination by means of odometry sensors, which count continuously from a known starting point Radumdrehungen and calculate the distance traveled based on a known wheel geometry. Models for slip behavior during start-up and braking, as well as up-to-date information about the condition of the wheels (for example in wet conditions) improve the calculation of the distance. Nevertheless, deviations arise repeatedly, e.g. with actively or passively communicating markings along the railway line whose coordinates are known (balises, RFID, etc.), corrected. However, this approach is comparatively complicated and expensive. It is also possible to carry out a coordinate determination of the rail vehicle by means of a global navigation satellite system (GNSS). However, this approach has the disadvantage that it works only limited in densely populated areas, in mountainous areas or in tunnels.

The object of the invention is to avoid the above-mentioned disadvantages and in particular to provide an efficient approach for determining the position of a rail vehicle.

This object is achieved according to the features of the independent claims. Preferred embodiments are in particular the dependent claims.

• – bei dem mittels einer Aufnahmeeinheit mindestens eine Aufnahme von der Strecke gemacht wird;
• – bei dem die Aufnahme mit zuvor abgespeicherten Daten von Landmarken verglichen wird,
• – bei dem bei zumindest einer vorgegebenen teilweisen Übereinstimmung der Aufnahme eine der Landmarken erkannt wird,
• – bei dem die Position des Schienenfahrzeugs anhand der erkannten Landmarke bestimmt wird.

To achieve the object, a method for determining the position of a rail vehicle is proposed,

• - Wherein by means of a recording unit at least one shot of the track is made;
• In which the recording is compared with previously stored data from landmarks,
• In which one of the landmarks is detected in at least one predetermined partial match of the recording,
• - in which the position of the rail vehicle is determined on the basis of the identified landmark.

Thus, it is advantageous that based on existing landmarks (e.g., also referred to as "natural" landmarks), the position of the rail vehicle can be determined and for this no extra active or passive markers need to be placed along the track. Furthermore, it is advantageous that due to the position determination already existing in the rail vehicle position determination can be checked and possibly also corrected. The approach is cheap and flexible, since a variety of such landmarks can be used to determine the position. In particular, the coordinates of such landmarks are known in many cases and do not have to be measured first.

A development is that the receiving unit on the rail vehicle, is arranged in particular with a receiving direction in the direction of travel of the rail vehicle.

The receiving unit may e.g. be aligned in the direction of travel; Also, the receiving unit may be arranged laterally on the rail vehicle.

In particular, a plurality of receiving units may be arranged on the rail vehicle, also at different locations (along) of the rail vehicle. Thus, it is also possible that a plurality of recording units deliver an image sequence that is prepared or processed accordingly.

Another development is that the previously stored data of landmarks are generated by means of at least one training session.

The training trip can be carried out especially for recording the landmarks and for storing the recordings associated with the landmarks. The training trip can also be part of a scheduled journey of a rail vehicle; In particular, the previously stored data of landmarks can be updated, adapted or checked in this way.

• – während der Trainingsfahrt Aufnahmen von den Landmarken gemacht werden und
• – die Aufnahmen aufbereitet und gespeichert werden.

In particular, it is a training that

• - during the training session recordings of the landmarks are made and
• - the recordings are prepared and stored.

Thus, it is a development that the recordings are prepared by at least one feature vector is determined per landmark and the at least one feature vector associated with the known position of the landmark and / or stored.

In particular, the feature vector may be stored on a route map at the position of the landmark, or the feature vector may be stored in a database together with the position of the landmark.

Further, it is a development that a plurality of training trips is performed and the previously stored data is averaged and / or adapted by means of the training trips.

In particular, it is possible that also a "normal" or regular drive, which serves for determining the position of the rail vehicle, is at least partially used as a training drive, by e.g. the recorded and optionally processed landmark or the feature vector determined from the image is used to average or adapt the stored data.

• – einer Odometrie des Schienenfahrzeugs,
• – einem globalen Navigationssatellitensystem,
• – mindestens einem zusätzlichen Sensor und/oder
• – aktiv oder passiv kommunizierenden Markierungen entlang der Strecke

genutzt wird, um die Position des Schienenfahrzeugs zu korrigieren und/oder zu validieren. A next development is that the position determination based on the recording unit in conjunction with

• - odometry of the rail vehicle,
• A global navigation satellite system,
• - At least one additional sensor and / or
• - actively or passively communicating markers along the route

is used to correct and / or validate the position of the rail vehicle.

• – eine Fotokamera,
• – eine Filmkamera,
• – eine Infrarotkamera,
• – ein Mikrofon,
• – ein Radargerät,
• – einen Sensor zur Ermittlung eines relativen Wegfortschritts.

An embodiment is that the receiving unit comprises at least one of the following components:

• - a photo camera,
• - a movie camera,
• - an infrared camera,
• - a microphone,
• A radar device,
• A sensor for determining a relative path progress.

In particular, the receiving unit can be made sensitive in certain wavelengths. Thus, especially at night or in the dark, e.g. in tunnels, usable recordings are made. A lighting unit can also be provided which illuminates a landmark along the route with a light in a predetermined wavelength range, so that a photograph can be taken of a recording unit sensitive in this wavelength range.

An alternative embodiment consists in that the recording comprises an image recording, in particular single images or moving images and / or a sound recording.

• – ein Bauwerk,
• – eine ortsfeste signaltechnische Einrichtung,
• – ein natürliches Landschaftsmerkmal,
• – ein Signal,
• – eine Weiche,
• – eine Tunneleinfahrt,
• – eine Tunnelausfahrt,
• – ein Gebäude,
• – ein Bahnübergang,
• – eine Gesteinsformation,
• – ein Landschaftszug,
• – eine Brücke.

A next embodiment is that the landmark is in particular one of the following components along the route:

• - a building,
• A stationary signaling device,
• - a natural landscape feature,
• A signal
• - a switch,
• - a tunnel entrance,
• - a tunnel exit,
• - a building,
• - a railroad crossing,
• - a rock formation,
• - a landscape train,
• - a bridge.

In particular, the stationary signaling equipment may comprise signals, points, milestones, level crossings, irregularities in the track bed or the rail.

The landmark can thus be seen, for example, next to the track, near the track or from the track. The landmark is, for example, a natural landmark, ie a phenomenological expression of the environment, which can be recorded by means of the recording unit. The landmark can also be a fixed feature.

The landmark is e.g. drawn on a route map, so that the absolute position of the landmark is known with high accuracy. If the rail vehicle passes the landmark, then the position of the rail vehicle can be determined with high accuracy at this time.

• – mittels der Aufnahmeeinheit mindestens eine Aufnahme von der Strecke gemacht wird,
• – die Aufnahme aufbereitet wird, indem ein Merkmalsvektor der Aufnahme, insbesondere einer aufgenommenen Szene, erstellt wird,
• – der Merkmalsvektor mit mindestens einem zuvor abgespeicherten Merkmalsvektor der Landmarke verglichen wird,
• – die Landmarke erkannt wird, falls der Merkmalsvektor und der mindestens eine zuvor abgespeicherte Merkmalsvektor der Landmarke zumindest teilweise übereinstimmen,
• – die Position des Schienenfahrzeugs anhand der Position der erkannten Landmarke bestimmt wird.

One embodiment is that

• At least one image is taken of the track by means of the recording unit,
• The recording is prepared by creating a feature vector of the recording, in particular a recorded scene,
• The feature vector is compared with at least one previously stored feature vector of the landmark,
• The landmark is detected if the feature vector and the at least one previously stored feature vector of the landmark at least partially match,
• - The position of the rail vehicle is determined by the position of the detected landmark.

A further development is that the recording is processed by retaining invariant characteristics of the landmark, regardless of minor changes in the recording.

An additional embodiment is that the position of the rail vehicle is determined based on the position of the detected landmark by a vehicle speed, a distance of the landmark from the rail vehicle, an angle of the recording, an enlargement of the recording and / or a processing speed for determining position are taken into account ,

The above statements apply to the method outlined below for training data, the rolling stock and the other categories of claim accordingly.

• – bei dem mittels einer Aufnahmeeinheit des Schienenfahrzeugs mindestens eine Aufnahme von einer Landmarke gemacht wird;
• – bei dem je Landmarke mindestens ein Merkmalsvektor bestimmt wird und der mindestens eine Merkmalsvektor mit der bekannten Position der Landmarke assoziiert und/oder gespeichert wird.

The above object is also achieved by means of a method for training data that can be used to determine the position of a rail vehicle,

• - Wherein by means of a receiving unit of the rail vehicle at least one shot of a landmark is made;
• In which at least one feature vector is determined per landmark and the at least one feature vector associated with the known position of the landmark and / or stored.

• – mit mindestens einer Aufnahmeeinheit,
• – mit mindestens einer Verarbeitungseinheit, die derart eingerichtet ist, dass
• – mittels der Aufnahmeeinheit mindestens eine Aufnahme von der Strecke machbar ist;
• – die Aufnahme mit zuvor abgespeicherten Daten von Landmarken vergleichbar ist,
• – bei zumindest einer vorgegebenen teilweisen Übereinstimmung der Aufnahme eine der Landmarken erkannt wird,
• – die Position des Schienenfahrzeugs anhand der erkannten Landmarke bestimmbar ist.

The above object is also achieved by means of a rail vehicle

• With at least one receiving unit,
• With at least one processing unit arranged such that
• - At least one shot of the track is feasible by means of the recording unit;
• The recording is comparable to previously stored data from landmarks,
• If one of the landmarks is detected in at least one predetermined partial match of the recording,
• - The position of the rail vehicle based on the detected landmark is determinable.

• – mit mindestens einer Aufnahmeeinheit,
• – mit mindestens einer Verarbeitungseinheit, die derart eingerichtet ist, dass
• – mittels der Aufnahmeeinheit des Schienenfahrzeugs mindestens eine Aufnahme von einer Landmarke machbar ist;
• – je Landmarke mindestens ein Merkmalsvektor bestimmbar ist und der mindestens eine Merkmalsvektor mit der bekannten Position der Landmarke assoziierbar und/oder speicherbar ist.

The problem is also solved by means of a rail vehicle

• With at least one receiving unit,
• With at least one processing unit arranged such that
• - At least one shot of a landmark is feasible by means of the receiving unit of the rail vehicle;
• - At least one feature vector is determined per landmark and the at least one feature vector with the known position of the landmark can be associated and / or stored.

• – Daten von Landmarken,
• – Merkmalsvektoren und/oder
• – Ergebnisse des Vergleichs der Aufnahme mit den zuvor abgespeicherten Daten von Landmarken

an das Schienenfahrzeug überträgt oder von diesem erhält. Furthermore, to solve the problem, a system is proposed with at least one rail vehicle as described herein and a control center (eg a computer or a computer network or network), which via a communication interface (eg, eg a mobile network in combination with a fixed network)

• - data of landmarks,
• - Feature vectors and / or
• - Results of the comparison of the recording with the previously stored data from landmarks

transmits to the rail vehicle or receives from this.

The solution presented here further comprises a computer program product which can be loaded directly into a memory of a digital computer, comprising program code parts which are suitable for performing steps of the method described here.

Furthermore, the above problem is solved by means of a computer-readable storage medium, eg of any memory, comprising computer-executable instructions (eg in the form of program code) suitable for the computer to perform steps of the method described herein.

The above-described characteristics, features, and advantages of this invention, as well as the manner in which they will be achieved, will become clearer and more clearly understood in connection with the following schematic description of exemplary embodiments which will be described in detail in conjunction with the drawings. For the sake of clarity, identical or identically acting elements may be provided with the same reference numerals.

Show it:

1 a schematic arrangement of a rail vehicle, which moves in the direction of travel along a route and has recording units, by means of which landmarks along the route can be recorded;

2 an exemplary schematic flowchart with steps of the method presented here;

3 an exemplary schematic flow diagram of a training as it is performed, for example, in the context of a training trip of a rail vehicle for creating training data, in particular feature vectors.

The approach presented here enables a flexible and, in particular, balise-free position determination of a rail vehicle. For example, in addition to GNSS receiver and route map position determination, safety may be enhanced by determining the position of the rail vehicle by means of a capture unit that provides recordings that are compared to previously stored landmark data along the route. If a landmark is detected, the location of the rail vehicle can also be determined via the stored position of the landmark.

Thus, the absolute position of the rail vehicle can be determined at certain times; it is also possible to validate or correct the position of the rail vehicle. In particular, it is possible to meet given safety requirements with a sufficient accuracy or tolerance.

The recording unit thus provides at least one additional sensor, by means of which stationary properties of the track can be detected. If these landmarks were previously detected (e.g., during at least one training run along the route) and, e.g. in the form of feature vectors together with their position, stored, then a recording of the landmark can be compared with the stored shape of the landmark. If there is a match or a partial match, the position of the landmark can be used to deduce the position of the rail vehicle. Preferably, the distance of the landmark to the rail vehicle, the distance, the speed of the rail vehicle and / or the processing time of the method are considered in order to determine the position of the rail vehicle based on the (recognized) landmark as accurately as possible at a certain time.

• – ein Signal,
• – eine Weiche,
• – eine Tunneleinfahrt,
• – eine Tunnelausfahrt,
• – ein Gebäude,
• – einen Bahnübergang,
• – eine Gesteinsformation,
• – einen Landschaftszug

handeln. The landmarks may, for example, at

• A signal
• - a switch,
• - a tunnel entrance,
• - a tunnel exit,
• - a building,
• - a railroad crossing,
• - a rock formation,
• - a landscape train

act.

The position of the landmark may e.g. stored in an electronic route map and possibly verified by (repeated) Streckenbefahrungen.

For example, after the at least one training session, the route map includes the original geo-coordinates and the sensor's position-independent location markers. Due to the diversity of the information in the route map, the system of GNSS receiver, recording unit, route map meets the given technical safety requirements.

The route map and / or the coordinates of the landmark may be part of a database that is accessed for determining the position of the rail vehicle.

In order to resolve ambiguities in the determination of landmarks or in their recognition, the evaluation of the data supplied by the acquisition unit can be extended by a regression analysis.

Optionally, a (further) sensor can be provided, on the basis of which a relative path progress is determinable; Thus, an additional plausibility of the landmark can be achieved. This relative displacement measurement can also be carried out on the basis of the recording unit.

In this case, it is advantageous for a route map to be extended by landmarks during a training phase (for example automatically). A locating method of the rail vehicle based thereon can be used e.g. In the absence of balises carry out a positionally safe position determination and thus meet the high safety requirements in rail transport.

A rail vehicle may e.g. at the front (aligned at the front in the direction of travel) be provided with at least one video camera as a recording unit that receives continuously or at predeterminable times the distance in the direction of travel. For example, images may be taken by the video camera at discrete times (e.g., every 2 seconds). Also, the video camera can continuously take pictures (e.g., as a movie at a predetermined number of frames per second). The video camera may be a digital or analog video camera. The images captured by the video camera may be local and / or central, e.g. stored on a video data server. In particular, the recordings, if necessary compressed or preprocessed, can be transmitted to the video data server. The transmission may be via a radio interface, e.g. a mobile communication network.

For example, the images are acquired using an algorithm, e.g. a model-based algorithm, prepared into scenes in which predetermined landmarks, as they occur along the railway line, can be detected. The landmarks are e.g. turnouts, signals, level crossings, bridges, etc.

From the landmarks the coordinates are known and e.g. in a route map stored on a train computer (OBU: "Onboard Unit"), e.g. in the form of a database or a route map, deposited.

For at least one training session, pictures of the landmarks can be taken. The recordings are prepared, for example, for later recognition (for example by means of an extraction of features of the landmarks by a wavelet algorithm) by determining feature vectors and storing them in a database, possibly together with the known coordinates of the respective landmark. The decision as to whether to include a landmark in the database can be made e.g. automatically (e.g., a signal can be automatically detected). It is also possible that certain landmarks were selected prior to the training trip, which are then specifically recorded, processed and stored in the form of feature vectors during the training session. Finally, an operator may also have certain landmarks, e.g. before or during the journey, which should be part of the database accordingly. Through repeated training runs, the created feature vectors, e.g. in different light conditions or weather conditions, adapted, e.g. be averaged in order to ensure as good as possible regardless of the environmental conditions a good recognisability.

In other words, during the at least one training session, shots can be taken of landmarks at the known positions of the landmarks. These recordings can be suitably processed and e.g. stored as feature vectors associated with the position of the landmark in a database and / or a route map. Thus, after training, a comparison of a shot (one image or a plurality of images) may be made to recognize one of the stored landmarks (eg, if a feature vector based on a current shot is similar to the landmark feature vector already stored in the database) ). Thus, based on the recording and the successfully performed comparison, the landmark can be recognized and the position associated with the landmark in the database or route map can be used to determine the position of the rail vehicle.

The feature vectors and the extraction algorithm are preferably chosen so that invariant properties of the landmark are maintained regardless of (minor) changes in the shot (e.g., soiled switch, railroad crossing animal, light and shadow effects, etc.).

1 shows a schematic arrangement of a rail vehicle 101 moving in the direction of travel 102 moved along a route. The rail vehicle 101 includes a calculator 103 (eg an OBU, a control unit, etc.), the data from here two example recording units 105 and 106 receives. The computer 103 can prepare such recordings, eg create scenes and / or determine feature vectors based on the recordings or the scenes and compare them with pre-recorded scenes and / or feature vectors. For this purpose, the calculator 103 locally to a database 104 access there, store recordings or feature vectors or read there existing data for comparison. In addition, the rail vehicle has 101 about a global Satellite Navigation System 108 as well as an odometry 109 for position determination. In this respect, an adjustment of the current position can also take place taking into account at least one of these two components.

The rail vehicle 101 has a communication interface 107 , eg in the form of a radio module or mobile communication device, with which via a radio interface 111 a connection to a wireless network 110 can be produced. Such a connection may continue via a wireless or wired interface 112 to a central office 113 (eg a computer, a network of computers or a computer network) exist, so that between the headquarters 113 and the rail vehicle 101 Data can be exchanged. The central office 113 can be distributed or centralized and have a variety of computers and / or data storage. By way of example, here is a database 114 displayed by the control panel. Database 114 For example, stores the feature vectors of the training rides in the form of a table or database or in the form of a route map.

With the example in 1 shown infrastructure can be a position determination of the rail vehicle 101 on the basis of the recording units 105 . 106 supplied data (images, sounds, films) with a comparison of previously recorded data (for example, in the database 104 and / or in the database 114 stored) whose position is known in each case are performed.

2 shows an exemplary schematic flowchart with steps of the method presented here. In one step 201 takes place by means of the receiving unit a recording in the direction of travel of the rail vehicle. The recording is processed, in particular a feature extraction is carried out based on scenes of the recording and feature vectors determined (see step 202 ). In one step 203 a comparison is performed, whereafter it is determined whether a currently determined feature vector is sufficiently similar to a feature vector stored in advance (such a comparison can also be performed for a number of feature vectors, a similarity measure may be cumulated in this case). If this is the case, it becomes a step 204 branched and determines the current position of the rail vehicle on the basis of the feature vector found to be sufficiently similar in comparison (or the position information associated with this feature vector). Optionally, then in one step 205 based on the determined position of the rail vehicle, a correction of the previously adopted (eg by means of the odometry) position of the rail vehicle performed or confirmed the correct position. Alternatively or additionally, in one step 206 the training data (ie the stored and considered for the comparison feature vectors) are adapted.

However, the comparison in step yields 203 in that the current feature vectors are not sufficiently similar, for example, becomes the step 201 branched.

3 shows an exemplary schematic flow diagram of a training (such as in the context of a training trip of a rail vehicle for creating training data, in particular feature vectors, is performed).

In one step 301 During the training drive, a landmark is recorded as a single image, image sequence or movie. In one step 302 a feature extraction of the recording takes place; this results in at least one feature vector. In one step 303 the at least one feature vector is stored or an adaptation of an already existing at least one feature vector is performed. Saving can be done in a database or in a route map.

In addition to the mentioned extraction algorithms, additional sensors can be used to achieve independence from environmental conditions. For example, normal cameras, infrared cameras and / or radars can be used. Such a radar could e.g. be implemented cost-effectively for the presently relevant short-range.

Also, regardless of environmental conditions (e.g., snow), the detection can be improved by having the camera focus on the detection of special objects (e.g., signals, kilometer tables) e.g. is optimized by a special orientation and / or focusing. Such special objects are e.g. Signals which have a high recognition value even in adverse environmental conditions and e.g. are arranged at a certain angle along the route.

The evaluation of the training rides can be interactive, i. with manual assistance to determine landmarks. Accordingly, the evaluation can also be fully automatic or at least partially automatic. In this case, an automated system can be provided which is self-learning. In particular, the result of multiple training runs can be used for automated detection.

In principle, it should be noted that the database with the feature vectors is running can be improved or supplemented. Thus, it is possible for a rail vehicle to receive data that is used on the one hand for position determination by means of the already existing feature vectors and on the other hand the recorded data are also used to improve the database in the manner described. Thus, even after an initial training, the database can be constantly monitored and improved, or the training can be continued (at least temporarily) even during normal operation.

Copies or parts of the database with the feature vectors for position determination are also accessible, for example, to the OBUs of the rail vehicles or provided on the rail vehicles. Alternatively or additionally, the database can be executed centrally; the database can e.g. be accessed by radio from the OBUs. In such an application, it is advantageous if predetermined maximum transmission times or response times can be maintained.

• (1) Die Aufnahme (Bilder oder Filme) wird normiert bzw. abgebildet auf z.B. Szenen, wie sie bei der Trainingsfahrt aufgenommen wurden oder wie sie im Anschluss an die Trainingsfahrt gespeichert wurden.
• (2) Es wird eine Merkmalsextraktion durchgeführt und ggf. ein Merkmalsvektor generiert nach demselben Vorgehen wie bei der Trainingsfahrt (z.B. kann dieselbe Wavelet-Transformation durchgeführt werden). Hierdurch werden vergleichbare Daten generiert, d.h. die Daten der aktuellen Fahrt werden mit den Daten der mindestens einen Trainingsfahrt vergleichbar.
• (3) Basierend auf dem Merkmalsvektor wird nach einem passenden oder einem am nächstbesten passenden Merkmalsvektor in der Datenbank gesucht. Hierbei kann ein schnelles Verfahren eingesetzt werden, das eine teilweise Übereinstimmung ausnutzt (PPM-Algorithmus).
• (4) Die Koordinaten des am besten passenden Treffers (der erkannten Landmarke) werden aus der Datenbank ausgelesen.
• (5) Ein Vorhaltewinkel der Kamera sowie eine Bearbeitungszeit für die Aufnahme, Erkennung und Zuordnung werden vorzugsweise so ausgelegt, dass auch bei Hochgeschwindigkeitszügen eine Auswertung abgeschlossen ist, wenn das Schienenfahrzeug die natürliche Landmarke überfährt.

An exemplary recognition algorithm running on the rail vehicle comprises the following steps:

• (1) The recording (images or films) is normalized or mapped to, for example, scenes as recorded during the training session or as stored after the training session.
• (2) A feature extraction is performed and if necessary a feature vector is generated according to the same procedure as during the training run (eg the same wavelet transformation can be performed). As a result, comparable data are generated, ie the data of the current trip are comparable to the data of at least one training session.
• (3) Based on the feature vector, a matching or closest matching feature vector is searched for in the database. Here, a fast method can be used that exploits a partial match (PPM algorithm).
• (4) The coordinates of the best matching hit (the detected landmark) are read from the database.
• (5) A lead angle of the camera and a processing time for the recording, detection and assignment are preferably designed so that even in high-speed trains, an evaluation is completed when the rail vehicle passes over the natural landmark.

The position information of the odometry and / or other sensors may be corrected based on the recognition of the landmark and associated coordinates.

Due to the corrected or updated or confirmed coordinates of the rail vehicle, a predetermined action can be performed. For example, the travel of the rail vehicle can be controlled accordingly.

It is also possible that an update of the feature vectors of the database is carried out on the basis of the images taken by the camera. Thus, e.g. the positions of landmarks, e.g. Changes due to rebuilding, recorded in this way and the database updated with the coordinates of the landmarks accordingly.

Moreover, it is possible that a change of the landmark is detected and e.g. a training trip is made after the conversion measure to update the database with the landmarks or the feature vectors corresponding to the landmarks.

Furthermore, it is an option to regularly or irregularly perform training rides to update the landmarks along the route. This can be done independently of known construction measures. These training rides have the advantage of having an early impact on landmarks, e.g. due to vandalism or weather: If, for example, a known landmark deposited in advance in the database with a feature vector is no longer recognized, this may be an indication that it has been removed or damaged.

For example, rail vehicles can transmit the recognized landmarks and / or the quality of recognition of the landmarks to a central office. The control panel determines therefrom e.g. a change of state of the landmark: "not recognizable or no longer available", "badly recognizable", "easily recognizable". Here, e.g. known weather conditions and / or times of the day (day, night, position of the sun, etc.) are taken into account. In particular, due to a large number of detections and / or a change in the quality of the recognition (normalized by weather condition and / or time of day), it can be determined automatically by the control center if a landmark is e.g. suddenly missing or (severely) damaged. In this case countermeasures can be initiated, e.g. the landmark can be repaired or redeployed.

Also, adaptive detection may be used to tailor feature vectors according to a landmark change. For example, a landmark, eg a building, may change color and / or shape over the years due to the effects of weather, so that recognition will be improved if this change is made in the feature vectors of the database is taken into account. If the change is greater than a predetermined threshold, in turn can be deduced to damage or theft of the landmark.

The approach presented here can be used individually or in combination with other location methods. With such a combination, one can increase the accuracy and reliability of determining the train position over the sole use of the conventional system, or reduce costs by partially reducing conventional positioning systems.

The solution presented here enables the evaluation of image information for position determination and / or for position correction of a rail vehicle. The advantage here is a model-based processing of images in image scenes, which are stored as feature vectors in a database. Furthermore, it is an advantage that invariant properties are stored in the feature vectors so that they are robust to minor changes and recognition of the landmark even in adverse and varying environmental conditions (day, night, rain, snow, sunshine, etc.) with high Security is possible.

For the recognition, an algorithm is preferably used which enables a correct recognition even with a partial match of a recording or the feature vector extracted therefrom with the feature vector stored in the database with high probability.

A particular advantage is that the existing "natural" landmarks along the route to additional artificial markers such as balises, RFID tags, or similar. at least partially waived. Phenomenological variations along the route (buildings, signals, tunnels, etc.), which are constant at least in the medium term, can thus be used efficiently to correct, determine or validate the position of the rail vehicle. Along each route, there are a variety of such natural landmarks whose position has been measured with high accuracy and is thus known. Such natural landmarks can thus be used efficiently for orientation for rail vehicles.

While the invention has been further illustrated and described in detail by the at least one embodiment shown, the invention is not so limited and other variations can be derived therefrom by those skilled in the art without departing from the scope of the invention.

QUOTES INCLUDE IN THE DESCRIPTION

This list of the documents listed by the applicant has been generated automatically and is included solely for the better information of the reader. The list is not part of the German patent or utility model application. The DPMA assumes no liability for any errors or omissions.

Cited non-patent literature

• http://www.itwissen.info/definition/lexikon/Wavelet-Transformation-WT-wavelet-transformation.html [0003]
• http://en.wikipedia.org/wiki/ Prediction_by_Partial_Matching [0004]

Claims (18)

Method for determining the position of a rail vehicle ( 101 ), - in which by means of a recording unit ( 105 . 106 ) at least one shot of the track is made ( 201 ); - in which the recording with previously stored data ( 104 . 114 ) is compared with landmarks ( 203 ), In which one of the landmarks is detected in at least one predetermined partial match of the recording ( 203 . 204 ), - in which the position of the rail vehicle is determined on the basis of the identified landmark ( 204 ). Method according to Claim 1, in which the receiving unit ( 105 . 106 ) on the rail vehicle ( 101 ), in particular with a receiving direction in the direction of travel ( 102 ) of the rail vehicle ( 101 ) is arranged. Method according to one of the preceding claims, in which the previously stored data are generated by landmarks by means of at least one training run. Method according to Claim 3, - in which recordings are made of the landmarks during the training journey ( 301 ) and - where the recordings prepared ( 302 ) and saved ( 303 ) become. Method according to Claim 4, in which the recordings are processed by determining at least one feature vector per landmark ( 302 ) and the at least one feature vector is associated and / or stored with the known position of the landmark ( 303 ). Method according to one of claims 4 or 5, wherein a plurality of training trips is performed and the previously stored data is averaged and / or adapted by means of the training trips. Method according to one of Claims 4 to 6, in which the training run is carried out during a regular journey. Method according to one of the preceding claims, in which the position determination on the basis of the recording unit is combined with - an odometry of the rail vehicle ( 109 ), - a global navigation satellite system ( 108 ), - at least one additional sensor and / or - actively or passively communicating markings along the track is used to correct the position of the rail vehicle and / or to validate. Method according to one of the preceding claims, wherein the receiving unit comprises at least one of the following components: - a photo camera, - a movie camera, - an infrared camera, - a microphone, A radar device, A sensor for determining a relative path progress. Method according to one of the preceding claims, in which the recording comprises an image recording, in particular single images or moving images and / or a sound recording. Method according to one of the preceding claims, in which the landmark is in particular one of the following components along the route: - a building, A stationary signaling device, - a natural landscape feature, A signal - a switch, - a tunnel entrance, - a tunnel exit, - a building, - a railroad crossing, - a rock formation, - a landscape train, - a bridge. Method according to one of the preceding claims, In which at least one image is taken of the track by means of the recording unit, In which the recording is prepared by creating a feature vector of the recording, in particular a recorded scene, In which the feature vector is compared with at least one previously stored feature vector of the landmark, In which the landmark is detected if the feature vector and the at least one previously stored feature vector of the landmark at least partially match, - In which the position of the rail vehicle is determined by the position of the detected landmark. The method of claim 12, wherein the shot is rendered by retaining invariant characteristics of the landmark, regardless of slight changes in the shot. Method according to one of the preceding claims, wherein the position of the rail vehicle is determined on the basis of the position of the detected landmark by a vehicle speed, a distance of the landmark from the Rail vehicle, an angle of the recording, an enlargement of the recording and / or a processing speed for determining the position are taken into account. Method for training data that can be used for determining the position of a rail vehicle, in which at least one image is taken of a landmark by means of a recording unit of the rail vehicle ( 301 ); In which at least one feature vector is determined per landmark ( 302 ) and the at least one feature vector is associated and / or stored with the known position of the landmark ( 303 ). Rail vehicle ( 101 ) - with at least one receiving unit ( 105 . 106 ), - with at least one processing unit ( 103 ), which is set up in such a way that - at least one recording of the route is feasible by means of the recording unit; - The recording with previously stored data from landmarks is comparable, - If at least a predetermined partial coincidence of recording one of the landmarks is detected, - The position of the rail vehicle based on the detected landmark can be determined. Rail vehicle ( 101 ) - with at least one receiving unit ( 105 . 106 ), - with at least one processing unit ( 103 ), which is arranged such that - by means of the receiving unit of the rail vehicle at least one recording of a landmark is feasible; - At least one feature vector is determined per landmark and the at least one feature vector with the known position of the landmark can be associated and / or stored.  System comprising at least one rail vehicle according to one of claims 16 or 17 and a control center, which has a communication interface - data of landmarks, - Feature vectors and / or - Results of the comparison of the recording with the previously stored data from landmarks transmits to the rail vehicle or receives from this.

Images